I was struck by one point Diane made about the future:
"But consistency in terms of what terms we choose to assign is not likely
to be as important in future as it has been in our current sharing
environment (which assumes that there is a "master record" that all will
use in the same way in a similar kind of system).... If we also consider
that in a more open data world there is no longer a real necessity to
consider multiple descriptions of the same item a problem (the information
can be aggregated in a number of ways, for instance),..."
I was struck by this, because one current trend in cataloging moving to
the so-called network level. That is, instead of each library maintaining
its own catalog and just taking copies of shared records to catalog and
maintain locally, we will truly share one big catalog and avoid doing
duplicative work at the local level. This goes beyond cooperative
cataloging as we now know it.
The example I am thinking of is WorldCat Local. It is a shared public
catalog, with relatively few tweaks and customization available to
individual libraries, and a common search and data platform that all
share. This is a case where the master record in fact becomes more
important than ever.
OCLC does not yet have a way to make institutional bib records (i.e., the
locally customized versions of master records) available for search and
discovery. Though they are pursuing ways of making some kinds of local
data available in local holdings records, this solution will not extend to
basic bibliographic fields in the master records in the foreseeable
future. And WorldCat Local cannot provide help like cross-references for
controlled vocabularies, partly because it is difficult to envision how to
implement that in a database that contains so many different controlled
vocabularies.
I don't live in a world where multiple descriptions and controlled
vocabularies are not be a problem. Not yet, anyway. So I understand any
nervousness about committing to a wholly new direction without being able
to see how it will really work. (And I have to say, I haven't been
following this thread very carefully at all. I just happened to see
Diane's remarks on multiple descriptions and was struck by the contrast
between her vision and our more immediate future.)
************
Diana Brooking (206) 685-0389
Cataloging Librarian (206) 685-8782 fax
Suzzallo Library [EMAIL PROTECTED]
University of Washington
Box 352900
Seattle WA 98195-2900
On Wed, 12 Nov 2008, Diane I. Hillmann wrote:
I think the report and this discussion about consistency is really
important for us all to think about. One of the ideas we'll need to
get used to as we contemplate changes to come is that data will come
from a variety of sources, and not all of those sources will be have the
same view of what's important or what needs to be consistent as we have
presently in our MARC-based shared cataloging systems. This suggests
that doing the thinking that the writers of the report have done (and
learning a bit from them) would be valuable to us as well.
I'd like to suggest here that one of the "consistency" issues brought up
below, the judgment of what is appropriate to assign for controlled
vocabularies and classification, is something that we've put a lot of
focus on in the past, in ways that we might not want to continue.
Consistency is still vitally important in terms of the VALUES that we
assign--this is where URIs help us, given that they imply that the URI
itself will reference an updated description of the subject term or
class number. But consistency in terms of what terms we choose to assign
is not likely to be as important in future as it has been in our current
sharing environment (which assumes that there is a "master record" that
all will use in the same way in a similar kind of system).
We can look at this situation in several ways: we can worry about the
"loss" of the traditional cataloger's common point of view that would
allow us to predict with some accuracy what another cataloger would
assign to a particular item (and feel empowered to judge whether that
assignment was "right" or "wrong"); or we can acknowledge the fact that
there are any number of possible points of view, and that users might be
advantaged by this variety (assuming that they might share one or more
of these more varied points of view). If we also consider that in a
more open data world there is no longer a real necessity to consider
multiple descriptions of the same item a problem (the information can be
aggregated in a number of ways, for instance), we start seeing
possibilities rather than just problems ahead. I believe it's really
important for us to consider whether our traditional emphasis on
consistency in all things is really a value we should carry forward, or
whether a more selective focus, emphasizing where consistency really
provides better results, serves us more.
As the report states, there are some opportunities here to use
machine-created metadata as well, particularly for the "factual" data
that machines can often do better, or at least more consistently, than
we can (determining the size of a digital resource, or the format, for
instance).
I've been accused of envisioning a world where catalogers are
superfluous, but that's really not what I think should or will happen.
What I envision, and hope for, is a world where machines can take over
the mind-numbing jobs that catalogers have always done, leaving the
higher level intellectual tasks that require human intervention to the
humans best able to do those tasks. This doesn't imply a world with
fewer catalogers but rather a world where a variety of people trained in
metadata practices will manage data describing a broader array of
resources, most likely not one-at-a-time. It's my expectation that most
of these people will have been catalogers in an earlier life.
It goes without saying that I think this approach will provide better
value for our users, as well--but that's another conversation.
Diane
Karen Coyle wrote:
Weinheimer Jim wrote:
This report concludes that consistency is not worthwhile, and I think
this paragraph (p. 17) sums it up quite well.
"At the other end of the spectrum is metadata of judgement. In
traditional cataloguing practice this is often called analytic
cataloguing; it involves the description of an item's 'aboutness' the
assigning of controlled vocabulary terms or classification numbers.
This sort of metadata is highly subjective and very expensive to
create since it usually requires subject and/or metadata experts
(cataloguers). Moreover, it is extremely dependent on the intended
audience of the metadata. One person who is describing an item for a
learning objects repository may assign one set of terms, and another
person describing the same item for a subject repository may describe
it another way. Same item. Different experts. Different audience.
Different metadata values. Given the expense and nature of this type
of content, we believe it is not feasible to expect consistency with
regard to metadata of judgement, except perhaps where it occurs in a
tightly controlled, narrow and consistent environment such as a
database of drug trials."
I interpret this statement differently than you do. Nowhere does the
report say that consistency is not "worthwhile" -- this is a study of
consistency, not the value of subject headings. Their conclusion, as you
quote above, is that consistency is unlikely across a broad spectrum of
metadata, especially between different communities. (Remember, this is a
report about repositories, not libraries, so we aren't talking MARC
records or AACR or even LCSH.) Since the question of consistency in this
report has to do with record sharing, they are pointing out that some
areas of the record are less likely to be consistent than others. They
aren't advocating AGAINST consistency, IMO.
Later, on p. 20 we read
"brief judgemental records are the domain of humans (and maybe
computers) but the content will often be thin and inconsistent;
fuller and judgemental records are too labour intensive for all but
clearly defined situations where ongoing costs have been accurately
estimated and benefits are considered to outweigh them; and computer
technology is not yet at a stage to replace human effort in this
regard."
So, what have we learned here? Subject analysis by catalogers is very
expensive (presumably because they make too much money), and they
don’t do their jobs very well. Therefore, subject analysis is not
worthwhile. Although retaining controlled vocabulary apparently is
worthwhile.
Again, no statement in the report about "worthwhile." That's your term.
An admission that it's expensive, that we don't have a good cost/benefit
ratio, and yet there is no way to automate this function. I think this
is no news at all.
I think you've mis-interpreted the report, Jim.
kc
--
-----------------------------------
Karen Coyle / Digital Library Consultant
[EMAIL PROTECTED] http://www.kcoyle.net
ph.: 510-540-7596 skype: kcoylenet
fx.: 510-848-3913
mo.: 510-435-8234
------------------------------------